Ion patent aims to verify authentic video in bytes

Ion Video has filed a patent application in Australia for technology that identifies whether a video is authentic or generated by artificial intelligence. The system is intended to address rising concerns about trust in video.

The ASX-listed company said the filing covers a method that authenticates video at the binary sample level, rather than by analysing on-screen pixels or relying on metadata. It creates a cryptographic record designed to show whether video content matches an original registered version.

Ion said the system checks content through what it calls an authentication policy engine. The engine combines several verification methods, including C2PA content credentials, camera sensor fingerprinting, codec forensics, AI detectors, studio attestations and capture-device registries, within a signed cryptographic record.

When a video is assembled and played back through its virtual video interface, the system can verify each binary sample against a registered hash in real time, according to the company. If a sample matches, it is treated as authentic. Any mismatch is flagged as unverified, modified or potentially synthetic.

Finbar O'Hanlon, Head of Innovation, said the aim was to create a way for video to demonstrate its authenticity at a more fundamental level than existing approaches.

"If the bytes match, the sample is authentic. If they do not, the sample is flagged as unverified, modified or potentially synthetic.

"What we have set out to do with this patent is give video a way to prove it is real. Our technology can verify content at the binary sample level using a number of methods, including cryptographic hashing, so the proof lives in the bytes themselves rather than in the pixels on the screen. It is designed to remain effective as generative AI advances, and to survive everyday changes to a file such as re-encoding and compression," said O'Hanlon.

Trust pressures

The filing comes as media groups, governments and technology companies face growing concern over how to distinguish manipulated or synthetic video from genuine footage. The spread of AI-generated content has made that task harder, particularly once files have been copied, compressed or redistributed across platforms.

O'Hanlon said many existing verification methods remain vulnerable because they rely on signals that can be removed or degraded during normal handling of video files.

"Existing approaches to this problem are largely reactive. Detection tools that analyse the decoded pixels of a video are engaged in a continual arms race with generative models, and each new detector can be defeated by the next generation of synthesis.

"Provenance metadata such as content credentials can be stripped or lost during re-encoding and redistribution. Invisible watermarks are fragile under common transformations such as compression and cropping, while hardware capture signatures verify only the single moment of capture. None of these methods combine multiple trust signals, and none operate at the binary sample level," he said.

He said Ion's approach differs by binding authentication to the internal structure of the video file and recording that information in a registry separate from the file itself. That setup allows organisations to define their own policies for which sources of trust must be met before content is accepted into the system, according to the company.

Those organisations could include studios, broadcasters, government agencies and large companies that need to verify video used in communications or internal workflows. Each can set its own policy rules for authentication sources before material is admitted, the company said.

Patent stack

Ion said the patent application completes a four-layer architecture within its Virtual Video Portfolio. It sits alongside three earlier patents and forms part of what the company describes as an end-to-end structure covering virtualise, record, govern and authenticate.

The company has linked that portfolio to a broader shift in how video is produced and distributed, moving away from fixed rendered files towards dynamic assembly of content. Ion said the patents together underpin infrastructure for that model.

O'Hanlon said the trust layer around video is under strain across sectors including news, entertainment, financial services, government and business communications.

"The trust layer that underpins video across news, entertainment, financial services, government and enterprise communications is increasingly under pressure," he said.